| I don't agree. First, a document containing an internal subset is not
| currently unconformant. Therefore UAs should be prepared to deal with
| internal subsets now. The question is how much.
I'd be happy if they did. But it would be a very large change.
For some folks perhaps.
| We can establish an HTML application convention that sanctions the
| interpretation of only certain declarations and ignores others.
Yes, I agree that in principle we can. However, your example
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN" [
<!NOTATION MyNotation PUBLIC "-//MyOrg//NOTATION MyNotation//EN">
<!ENTITY MyRandomData SYSTEM "http://foo.com/random.dat"
NDATA MyNotation>
<!ENTITY MyPCData SYSTEM "http://foo.com/pcdata.htm">
]>
suggests that the convention would be "attend to notation and entity
declarations, but ignore element declarations."
I don't have any SGML tools that will do that (so far as I know),
but at the moment I can vet all my docs with my SGML tools. I don't
want that to change.
I'm not talking about making SGML tools more restrictive, but making HTML
UAs more inclusive w.r.t. SGML.
I have to confess that it is relatively easy for me to call for this
approach since our (Stonehand) HTML Viewer employs an SGML-based parser
which does support internal subsets but which has been made somewhat more
restrictive in the sense that it does not permit the declaration or
redeclaration of element types, attribute lists, shortrefs, linktypes,
etc. [If it encounters one of these, it ignores it and emits a warning.]
What it does support is full use of notation and general entity declarations.
I therefore don't think it unreasonable to propose the use of external
entity declarations to accomplish the desired goal of data inclusion.
Glenn